@@ -0,0 +1,25 @@
+function m = lightsheetMethod()
+%LIGHTSHEETMETHOD The method term that marks a migrated lightsheet pyramid.
+%
+%   M = ndi.vintage.lightsheetMethod()
+%
+%   M is struct('node', 'FBbi:00000369', 'name', 'selective plane
+%   illumination fluorescence microscopy') -- light-sheet microscopy (checked
+%   on OLS). DID-matlab +migrators_j/lightsheet_zarr_pyramid.m sets it as
+%   `interaction.method` on the intensity observation it makes of EVERY v1
+%   lightsheetZarrPyramid, OME-Zarr and SmartSPIM alike (Jess 2026-10-15). It
+%   is what says "lightsheet pyramid" rather than "some fluorescence
+%   measurement", so ndi.vintage.lightsheetPyramidDocs finds migrated pyramids
+%   by its node. It must stay the migrator's, character for character;
+%   ndi.unittest.vintage.TestLightsheetMigrated pins it against the migrator's
+%   output when DID-matlab's migrators are on the path.
+%
+%   (Until 2026-10-15 the marker was the name of the builder `software`
+%   entity the observation's software_id names. That entity is still
+%   written, as the record of what built the pyramid; it no longer marks it.)
+%
+%   See also: ndi.vintage.lightsheetPyramidDocs.
+
+m = struct('node', 'FBbi:00000369', ...
+    'name', 'selective plane illumination fluorescence microscopy');
+end
@@ -26,11 +26,21 @@
 %                      (LEVELRECS, ndi.vintage.lightsheetLevels), which the
 %                      same pass completed from the pyramid's units and
 %                      channel names.
-%     label, pyramid_name, pyramid_type
-%                      the migration keeps these ONLY as words in the store
-%                      body's `description` ('label "..."', 'NGFF multiscales
-%                      name "..."', '... type "..."'), so they are parsed
-%                      back out of it; '' when there is no store body.
+%     label            the observation's `statement.notes` (Jess
+%                      2026-10-15: the label stays there, permanently). When
+%                      the batch pass did not run, the notes also carry, after
+%                      a newline and '-- store body: ', the text the pass
+%                      would have moved to the store body; only the part
+%                      before the LAST such separator is the label. The
+%                      separator is DID-matlab's lightsheet_zarr_pyramid.m /
+%                      foldGenericFiles.m NOTES_SEPARATOR, character for
+%                      character.
+%     pyramid_name     parsed from the store body's `description`
+%                      ('NGFF multiscales name "..."'); '' when there is no
+%                      store body (or no SESSION).
+%     pyramid_type     '' always: the migration DROPS it (Jess 2026-10-15; it
+%                      repeats the levels' summary statistic, and the v1
+%                      docs say not to trust it).
 %     pipeline_version ''. It became a `version` text_assertion about the
 %                      software entity, not read here.
 %     byte_order       level 0's, if given.
@@ -53,6 +63,8 @@
     'shape_level0', [], 'voxel_size_level0', [], 'voxel_size_units', '', ...
     'translation_level0', [], 'dtype', '', 'pipeline_version', '', 'byte_order', '');
 
+p.label = labelOf(charOr(ndi.v2.blockOf(props, 'statement', 'notes', '')));
+
 p.dtype = charOr(ndi.v2.blockOf(props, 'value', 'source_data_type'));
 if isempty(p.dtype)
     p.dtype = charOr(ndi.v2.blockOf(props, 'value', 'data_type'));
@@ -66,10 +78,7 @@
         sp = stores{k}.document_properties;
         d = blockValue(sp, 'description');
         if ~isempty(d)
-            d = charOr(d);
-            p.label = quoted(d, 'label');
-            p.pyramid_name = quoted(d, 'NGFF multiscales name');
-            p.pyramid_type = quoted(d, 'NGFF multiscales type');
+            p.pyramid_name = quoted(charOr(d), 'NGFF multiscales name');
         end
         keys = blockValue(sp, 'keys');
         if ~isempty(keys) && isempty(frame)
@@ -112,6 +121,18 @@
 end
 end
 
+function label = labelOf(notes)
+% The label: NOTES up to the LAST separator DID-matlab's migrator writes
+% before the text its store-record pass moves out (present only when that pass
+% did not run), else all of NOTES.
+SEPARATOR = sprintf('\n-- store body: ');
+label = notes;
+at = strfind(notes, SEPARATOR);
+if ~isempty(at)
+    label = notes(1:at(end) - 1);
+end
+end
+
 function s = quoted(text, label)
 s = '';
 tok = regexp(text, [regexptranslate('escape', label) ' "([^"]*)"'], 'tokens', 'once');
 
@@ -8,45 +8,41 @@
 %
 %   HOW A MIGRATED PYRAMID IS RECOGNISED. DID-matlab
 %   +migrators_j/lightsheet_zarr_pyramid.m makes each v1 pyramid an intensity
-%   `observation` (its id kept) whose software_id names an `entity` of type
-%   software called ndi.vintage.lightsheetBuilderName() -- 'NDI lightsheet
-%   pyramid builder (ndi.fun.doc.lightsheet.makePyramid)' (Jess
-%   2026-10-14, lightsheet point 5). Nothing else on the observation says
-%   "lightsheet": its class is `observation`, its value kind `intensity` and
-%   its variable NCIT:C184706 'Fluorescence Intensity', all of which any
-%   fluorescence measurement shares. So a migrated pyramid is an observation
-%   whose software is THAT entity, found in two steps (the entity by name,
-%   then the observations pointing at it) because a query sees one document
-%   at a time.
+%   `observation` (its id kept) whose `interaction.method` is
+%   ndi.vintage.lightsheetMethod() -- FBbi:00000369 'selective plane
+%   illumination fluorescence microscopy' (Jess 2026-10-15). Its class
+%   (`observation`), value kind (`intensity`) and variable (NCIT:C184706
+%   'Fluorescence Intensity') are shared by any fluorescence measurement; the
+%   method is what says light-sheet. So a migrated pyramid is an observation
+%   whose method node is that term, found in ONE query on the field path
+%   `interaction.method.node` (`interaction` declares `method`, an
+%   ontology_term {node, name}; ndi.v2.fieldQuery also asks the block's name
+%   before 2026-10-08, `subject_interaction`).
+%
+%   (Until 2026-10-15 it was found through the builder `software` entity's
+%   name. The entity is still written; it no longer marks the pyramid.)
 %
 %   This is why lightsheet has no ndi.vintage.map row. ndi.vintage.entryFor
 %   keys on the class name alone, and `observation` / `sampled_body` are every
 %   V_eta statement and array: a row would reroute the edge and field reads
 %   ndi.vintage.edge / field do for EVERY observation in a session (e.g. the
 %   time-reference reads in ndi.file.navigator.bodies), not just pyramids.
 %
-%   The software entity is matched in both shapes did-schema has used: an
-%   `entity` with name (current) or a `software` document with name (before
-%   2026-10-08).
-%
-%   See also: ndi.vintage.lightsheetLevelDocs, ndi.vintage.lightsheetPyramid.
+%   See also: ndi.vintage.lightsheetMethod, ndi.vintage.lightsheetLevelDocs,
+%             ndi.vintage.lightsheetPyramid.
 
-BUILDER = ndi.vintage.lightsheetBuilderName();
+M = ndi.vintage.lightsheetMethod();
 
 docs = session.database_search(ndi.query('', 'isa', 'lightsheetZarrPyramid'));
 docs = reshape(docs, 1, []);
 
-sw = session.database_search(ndi.query('entity.name', 'exact_string', BUILDER, '') | ...
-    ndi.query('software.name', 'exact_string', BUILDER, ''));
-seen = {};
-for i = 1:numel(sw)
-    obs = session.database_search(ndi.v2.isaQuery('observation') & ...
-        ndi.query('', 'depends_on', 'software_id', sw{i}.id()));
-    for j = 1:numel(obs)
-        id = obs{j}.id();
-        if any(strcmp(seen, id)), continue; end
-        seen{end+1} = id; %#ok<AGROW>
-        docs{end+1} = obs{j}; %#ok<AGROW>
-    end
+obs = session.database_search(ndi.v2.isaQuery('observation') & ...
+    ndi.v2.fieldQuery('interaction.method.node', 'exact_string', M.node));
+seen = cellfun(@(d) d.id(), docs, 'UniformOutput', false);
+for j = 1:numel(obs)
+    id = obs{j}.id();
+    if any(strcmp(seen, id)), continue; end
+    seen{end+1} = id; %#ok<AGROW>
+    docs{end+1} = obs{j}; %#ok<AGROW>
 end
 end
@@ -7,19 +7,22 @@
 %   did2.build.key: a regular axis's origin and spacing are {value,
 %   source_value} pairs) and the store-record
 %   half of +did2/+convert/foldGenericFiles.m, DID-matlab
-%   claude/ecstatic-pascal-md91fw at 000aeb8), AFTER the batch pass:
+%   claude/ecstatic-pascal-md91fw at f363d98), AFTER the batch pass:
 %
 %     pyramid  an `observation` (value kind `intensity`) with the v1 pyramid's
 %              id; entity_id = the subject, instrument_id = the element,
 %              software_id = the builder `entity`; value.data_type 'uint16',
 %              value.source_data_type '<u2' (the v1 dtype), data_body true;
-%              variable NCIT:C184706 'Fluorescence Intensity'.
-%     software an `entity` of type software named
-%              ndi.vintage.lightsheetBuilderName().
+%              variable NCIT:C184706 'Fluorescence Intensity'; method
+%              ndi.vintage.lightsheetMethod() (FBbi:00000369, Jess
+%              2026-10-15), the mark ndi.vintage.lightsheetPyramidDocs finds;
+%              notes = the v1 label (Jess 2026-10-15).
+%     software an `entity` of type software named for the builder (the
+%              migrator's BUILDER; no longer the finder's mark).
 %     store    the v1 fileReference as an UNHELD `opaque_body` owned by the
 %              observation, format EMPTY:0000259 OME-Zarr, carrying the
 %              full-resolution keys and a description holding the pyramid's
-%              label / NGFF name / type.
+%              NGFF multiscales name (no label, no type: Jess 2026-10-15).
 %     level    a `sampled_body` with the v1 level's id, owner_id = the
 %              observation; keys t,c,z,y,x (t in second, c named by channel,
 %              z/y/x converted to meter with the micrometre source kept);
@@ -45,7 +48,9 @@
             s.depends_on = struct('name', {'entity_id', 'instrument_id', 'software_id'}, ...
                 'document_id', {'subj_ls_1', 'elem_scope_1', softwareId});
             s.statement = struct('variable', ...
-                struct('node', 'NCIT:C184706', 'name', 'Fluorescence Intensity'));
+                struct('node', 'NCIT:C184706', 'name', 'Fluorescence Intensity'), ...
+                'notes', 'brain 7 cleared');
+            s.interaction = struct('method', ndi.vintage.lightsheetMethod());
             s.value = struct('data_type', 'uint16', 'source_data_type', '<u2', ...
                 'data_body', true);
         end
@@ -54,7 +59,7 @@
             s = ndi.unittest.vintage.LightsheetFixtures.etaDoc('entity', {'base'}, id);
             s.depends_on = struct('name', {}, 'document_id', {});
             s.entity = struct('type', struct('node', '', 'name', 'software'), ...
-                'name', ndi.vintage.lightsheetBuilderName());
+                'name', 'NDI lightsheet pyramid builder (ndi.fun.doc.lightsheet.makePyramid)');
         end
 
         function s = store(id, ownerId)
@@ -66,8 +71,8 @@
                 'format', struct('node', 'EMPTY:0000259', 'name', 'OME-Zarr'), ...
                 'filename', 'brain7.ome.zarr', ...
                 'description', ['OME-Zarr store, not held in the database; ' ...
-                    'label "brain 7 cleared"; NGFF multiscales name "mean"; ' ...
-                    'NGFF multiscales type "mean"; .zattrs MD5 0123456789abcdef'], ...
+                    'NGFF multiscales name "mean"; recorded by location, not ' ...
+                    'ingested; .zattrs MD5 0123456789abcdef'], ...
                 'redundant', false);
         end
 
 
@@ -12,7 +12,7 @@
 %   ndi.session.dir, and asks the same readers.
 %
 %   NEEDS DID-matlab's lightsheet migrators (claude/ecstatic-pascal-md91fw,
-%   000aeb8 and later) and the composition did-schema they target, so
+%   f363d98 and later: the method and the label on notes, Jess 2026-10-15) and the composition did-schema they target, so
 %   test-import-v2.yml runs it on the `next` leg only. It is not filtered on
 %   `current`; it is not selected there.
 %
@@ -60,27 +60,66 @@ function migrateAndWrite(testCase)
 
     methods (Test)
 
-        function testTheBuilderNameIsTheMigrators(testCase)
+        function testTheMethodIsTheMigrators(testCase)
             % ndi.vintage.lightsheetPyramidDocs finds migrated pyramids by
-            % this one string; it must be the migrator's, character for
-            % character.
-            names = {};
+            % this method node (Jess 2026-10-15), on the path
+            % interaction.method.node; it must be the migrator's, character
+            % for character, and at that path.
+            obs = [];
             for k = 1:numel(testCase.Out)
-                d = testCase.Out{k};
-                if strcmp(d.document_class.class_name, 'entity') || ...
-                        strcmp(d.document_class.class_name, 'software')
-                    n = ndi.v2.blockOf(d, 'software', 'name', '');
-                    if ~isempty(n), names{end+1} = char(n); end %#ok<AGROW>
-                end
+                if strcmp(testCase.Out{k}.base.id, 'pyr_ls_1'), obs = testCase.Out{k}; end
             end
-            testCase.verifyTrue(any(strcmp(names, ndi.vintage.lightsheetBuilderName())), ...
-                sprintf('software entities in the migration: %s', strjoin(names, ' | ')));
+            testCase.assertNotEmpty(obs, 'the pyramid''s observation');
+            m = ndi.v2.blockOf(obs, 'interaction', 'method', []);
+            if iscell(m), m = m{1}; end
+            testCase.assertTrue(isstruct(m) && isfield(m, 'node'), ...
+                'the migrated pyramid carries interaction.method');
+            M = ndi.vintage.lightsheetMethod();
+            testCase.verifyEqual(char(m.node), M.node);
+            testCase.verifyEqual(char(m.name), M.name);
+        end
+
+        function testTheLabelStaysOnTheObservation(testCase)
+            % Jess 2026-10-15: the label is the observation's notes after the
+            % batch pass, and the store body's description holds the NGFF
+            % multiscales name only.
+            d = testCase.Session.database_search(ndi.query('base.id', 'exact_string', 'pyr_ls_1'));
+            testCase.assertNumElements(d, 1);
+            testCase.verifyEqual(char(ndi.v2.blockOf(d{1}.document_properties, ...
+                'statement', 'notes', '')), 'brain 7 cleared');
+            [p, vintage] = ndi.vintage.lightsheetPyramid(d{1}, testCase.Session);
+            testCase.verifyEqual(vintage, 'V_eta');
+            testCase.verifyEqual(p.label, 'brain 7 cleared');
+            testCase.verifyEqual(p.pyramid_name, 'mean');
+            testCase.verifyEqual(p.pyramid_type, '', 'pyramid_type is dropped');
+        end
+
+        function testWithoutThePassTheReaderStillGetsTheLabel(testCase)
+            % A bare v1_to_v2 call leaves the multiscales name after the
+            % migrator's separator in notes; the reader's separator must be
+            % the same literal, so the label it reads is the label alone.
+            out = did2.convert.v1_to_v2({ndi.unittest.vintage.TestLightsheetMigrated.v1Pyramid()}, ...
+                'Validate', true, 'TargetVersion', 'V_eta');
+            testCase.assertEmpty(out.quarantine, ...
+                ndi.unittest.vintage.TestLightsheetMigrated.reasons(out));
+            obs = [];
+            for k = 1:numel(out.migrated)
+                d = out.migrated{k}.toStruct();
+                if strcmp(d.base.id, 'pyr_ls_1'), obs = d; end
+            end
+            testCase.assertNotEmpty(obs);
+            notes = char(ndi.v2.blockOf(obs, 'statement', 'notes', ''));
+            testCase.verifySubstring(notes, 'NGFF multiscales name "mean"', ...
+                'without the pass the name is still on the notes');
+            p = ndi.vintage.lightsheetPyramid(obs);
+            testCase.verifyEqual(p.label, 'brain 7 cleared');
         end
 
         function testTheMigratedPyramidIsFoundAndListed(testCase)
             docs = ndi.vintage.lightsheetPyramidDocs(testCase.Session);
             ids = cellfun(@(d) d.id(), docs, 'UniformOutput', false);
-            testCase.verifyEqual(ids, {'pyr_ls_1'}, 'the pyramid keeps its id');
+            testCase.verifyEqual(ids, {'pyr_ls_1'}, ...
+                'the pyramid keeps its id, and is found by its method');
             rows = ndi.gui.app.LightsheetZarrManager.pyramidRows(testCase.Session);
             testCase.verifyNumElements(rows, 1);
             testCase.verifyEqual(rows(1).subjectID, 'subj_ls_1');
 
@@ -41,8 +41,14 @@ function needV2(testCase)
             mkdir(bytes);
             testCase.Member0 = F.tempFile(bytes, 'uid_level0_chunk0', uint8(1:8));
             m1 = F.tempFile(bytes, 'uid_level0_chunk1', uint8(9:16));
+            % A DECOY: an intensity observation with the same builder
+            % software but no light-sheet method. The finder keys on the
+            % method (Jess 2026-10-15), so it must NOT be listed.
+            decoy = F.pyramid('obs_not_lightsheet', 'sw_ls_1');
+            decoy.interaction = struct('method', struct('node', '', 'name', ''));
+            decoy.statement.notes = '';
             docs = {F.software('sw_ls_1'), ...
-                F.pyramid('pyr_ls_1', 'sw_ls_1'), ...
+                F.pyramid('pyr_ls_1', 'sw_ls_1'), decoy, ...
                 F.store('fref_store_1', 'pyr_ls_1'), ...
                 F.level('lvl_0', 'pyr_ls_1', 1, '', {testCase.Member0, m1}), ...
                 F.level('lvl_max1', 'pyr_ls_1', 2, 'max', {}), ...
@@ -61,7 +67,8 @@ function testTheMigratedPyramidIsFound(testCase)
             docs = ndi.vintage.lightsheetPyramidDocs(testCase.Session);
             ids = cellfun(@(d) d.id(), docs, 'UniformOutput', false);
             testCase.verifyEqual(ids, {'pyr_ls_1'}, ...
-                'exactly the observation whose software is the pyramid builder');
+                ['exactly the observation whose method is light-sheet ' ...
+                 'microscopy, not every observation of the builder software']);
         end
 
         function testLevelTableReadsTheMigratedLevels(testCase)
@@ -113,7 +120,8 @@ function testTheManagerListsTheMigratedPyramid(testCase)
             rows = ndi.gui.app.LightsheetZarrManager.pyramidRows(testCase.Session);
             testCase.verifyNumElements(rows, 1);
             testCase.verifyEqual(rows(1).id, 'pyr_ls_1');
-            testCase.verifyEqual(rows(1).label, 'brain 7 cleared');
+            testCase.verifyEqual(rows(1).label, 'brain 7 cleared', ...
+                'the label is the observation''s notes (Jess 2026-10-15)');
             testCase.verifyEqual(rows(1).subjectID, 'subj_ls_1');
             testCase.verifyEqual(rows(1).nLevels, 3);
             testCase.verifyEqual(rows(1).reduction, 'max/mean');